The other talk that I found interesting today was one by Mike Erwin on the various trade-offs incurred when invoking SQL parallel queries. The sub-text for this talk was, "Are RAC and performance mutually exclusive?" (it's great to see people asking questions like this). The actual content of the talk was it little bit too much focused on knob turning for my taste, since I don't use RAC and I also prefer to look for general principles. Nonetheless, I liked Mike's presentation style and I spoke with him afterwards. It turns out that he has RAC scalability data which he is prepared to share with me, so I'm looking forward to the possibility of throwing my universal scalability model at those data in the not too distant future. Stay tuned!
Unfortunately, the last talk I was planning to attend, "Queueing Theory Analysis of Statspack Reports Obtained from Database Load Tests" was cancelled due to the author being ill.
Finally, I met with Gerwin Hendriksen because he asked my to attend his presentation on Monday and critique it. Gerwin was very proud of his little statistical discovery (as he should be). What neither of us knew was that I not only knew about the technique but had used something very similar years earlier in a different performance context (see Chap. 8 of my GCaP book). Therefore, as you can see, Gerwin and I had a very "intense" and "scientific" chat about his presentation over a couple of glasses of Australian Shiraz (compliments of Performance Dynamics). During his presentation on Monday, I suddenly realized that Gerwin's approach was actually more general than he himself was claiming, but he had not noticed it. During his talk, when I spontaneously felt compelled to point out that one of his assumptions could be relaxed to make his technique even more encompassing, I think some of the audience thought I was attacking him. Certainly, he was broadsided by my observation [I'm good at that :-)] but as the photo shows, no enmity was involved.
No comments:
Post a Comment